Automatic Speech Recognition on VoxPopuli
5.55WERAudio Flamingo 3
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Audio Flamingo 3Backbone=Audio Flamingo 3, Method=Baseline2026.06 | 5.55 | — | |
| Qwen2.5-Omni-7B + CoATBackbone=Qwen2.5-Omni-7B, Method=CoAT2026.06 | 5.56 | — | |
| Canary-1b-flashArchitecture Type=AR, Params (B)=1.0, RTFx=29.25, Avg. WER=7.22026.02 | 5.6 | — | |
| Audio Flamingo 3 + CoATBackbone=Audio Flamingo 3, Method=CoAT2026.06 | 5.7 | — | |
| Qwen2.5-Omni-7BBackbone=Qwen2.5-Omni-7B, Method=Baseline2026.06 | 5.8 | — | |
| Mixed Training + Length Aug + Reduced TFLength Augmentation=true, Reduced Teacher Forcing=true2026.04 | 5.95 | — | |
| Phi-4-multimodalArchitecture Type=AR, Params (B)=5.6, RTFx=4.85, Avg. WER=6.72026.02 | 6 | — | |
| MDM-ASRArchitecture Type=NAR, Params (B)=1.0, RTFx=46.81, Avg. WER=6.92026.02 | 6 | — | |
| CrisperWhisper2026.04 | 6.03 | — | |
| Mixed Training + Length AugLength Augmentation=true2026.04 | 6.05 | — | |
| Mixed Training + Length Aug + Timestamp RegLength Augmentation=true, Timestamp Regularization=true2026.04 | 6.15 | — | |
| Phi-4-MultimodalType=AR2026.01 | 6.3 | — | |
| ASR-only Baseline2026.04 | 6.31 | — | |
| Qwen2.5-Omni-3B# Total Params=5B2025.11 | 6.59 | — | |
| Parakeet-CTCArchitecture Type=NAR, Params (B)=1.1, RTFx=120.20, Avg. WER=8.32026.02 | 6.6 | — | |
| MinMoType=AR2026.01 | 6.7 | — | |
| Mixed Training2026.04 | 7.02 | — | |
| Qwen2-Audio + CoATBackbone=Qwen2-Audio, Method=CoAT2026.06 | 7.03 | — | |
| Qwen2-AudioType=AR2026.01 | 7.1 | — | |
| DIFFUSPEECHType=Diff.2026.01 | 7.1 | — | |
| Qwen2-AudioArchitecture Type=AR, Params (B)=8.4, RTFx=4.18, Avg. WER=8.32026.02 | 7.1 | — | |
| Voxtral-MiniArchitecture Type=AR, Params (B)=3.0, RTFx=12.19, Avg. WER=8.02026.02 | 7.1 | — | |
| Qwen-Audio2026.04 | 7.22 | — | |
| OWSM-v3.1Architecture Type=AR, Params (B)=1.0, RTFx=15.52, Avg. WER=10.32026.02 | 8.4 | — | |
| OWSM-CTCArchitecture Type=NAR, Params (B)=1.0, RTFx=132.54, Avg. WER=11.42026.02 | 8.6 | — | |
| DraxArchitecture Type=NAR, Params (B)=1.2, RTFx=11.32, Avg. WER=9.22026.02 | 8.6 | — | |
| MoshiType=AR2026.01 | 8.8 | — | |
| Qwen3-ASR-1.7BParameters=1.7B2026.01 | 9.15 | — | |
| Llama-Omni2Type=AR2026.01 | 9.5 | — | |
| Whisper-large-v3Architecture Type=AR, Params (B)=1.5, RTFx=12.83, Avg. WER=8.52026.02 | 9.5 | — | |
| LFM2-Audio-1.5B# Total Params=1.5B2025.11 | 9.93 | — | |
| Qwen3-ASR-0.6BParameters=0.6B2026.01 | 9.96 | — | |
| GPT-4o-Transcribe2026.01 | 10.29 | — | |
| Gemini-2.5-Pro2026.01 | 11.36 | — | |
| Whisper-large-v3-turbo# Total Params=1.5B2025.11 | 11.87 | — | |
| Qwen2-AudioBackbone=Qwen2-Audio, Method=Baseline2026.06 | 11.9 | — | |
| Whisper-large-v32026.01 | 12.05 | — | |
| Doubao-ASR2026.01 | 12.12 | — | |
| SpiritLMType=AR2026.01 | 14.3 | — | |
| SpeechGPTType=AR2026.01 | 18.2 | — | |
| Whisper-CDModel=Large-v32026.03 | 19.86 | — | |
| Whisper-CDModel=Large-v3-Turbo2026.03 | 25.71 | — | |
| Whisper BaselineModel=Large-v3-Turbo2026.03 | 30.63 | — | |
| Whisper BaselineModel=Large-v32026.03 | 44.95 | — | |
| Canary-1B-v22026.05 | — | 6.2 | |
| Doubao-LLM ASR2026.05 | — | 7.14 | |
| Gemini-3-Flash2026.05 | — | 7.74 | |
| GPT-4o-trans.2026.05 | — | 7.02 | |
| Kimi-Audio-7B2026.05 | — | 8.15 | |
| MEGA-ASR2026.05 | — | 7.44 | |
| MEGA-ASRrouter=true2026.05 | — | 6.26 | |
| Parakeet-TDT-0.6B-v32026.05 | — | 6.11 | |
| Qwen2.5-Omni-7B2026.05 | — | 6.02 | |
| Qwen3-ASR-1.7B2026.05 | — | 6.25 | |
| Step-Audio-2-mini2026.05 | — | 7.43 | |
| Voxtral-Mini-3B-25072026.05 | — | 7.08 | |
| Whisper Large-v32026.05 | — | 9 |